SciMaster

SciMaster is the first general-purpose research agent built for scientific work. It grew out of a joint project between Shanghai Jiao Tong University, the Shanghai Algorithm Innovation Institute, and DP Technology, the company behind the Bohrium science platform. Instead of answering with a few loose snippets, it plans a research question the way a grad student would: it breaks the problem into sub-tasks, searches papers and the open web, runs calculations, and hands back a written report. Many people now describe this kind of AI research work as "AI for science," and SciMaster is a direct attempt to make it usable.

The idea behind it is simple. Scientific literature has exploded while the human reading speed hasn't, and modern problems often need tools from several fields at once. SciMaster tries to close that gap by chaining many steps together in one place. It draws on Bohrium's library of roughly 170 million research papers and can call scientific software for tasks like molecular simulation and DFT calculations.

The catch is that it's built for people with real research questions. It isn't a chat toy, and it won't replace domain judgment. You still need to know your field well enough to read the output critically. Access has also been limited in waves, with early users joining through invitation codes. For everyday writing or quick factual lookups, a general chatbot is faster and cheaper.

Getting Started

1. Create a Bohrium account and open the SciMaster agent from the Bohrium workspace.
2. Type a precise scientific question, such as a review topic or a method comparison you need mapped out.
3. Start the run and watch SciMaster split the problem into sub-tasks. You can pause the plan and edit the chain of thought before it continues.
4. Review the generated report, check the cited sources, and export or share it as needed.